The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash(agent-browser:*) tool to execute a wide variety of browser automation commands. This includes high-risk operations such as agent-browser eval for executing arbitrary JavaScript in the page context and agent-browser network route for intercepting or mocking network requests.
[DATA_EXFILTRATION]: Several commands allow for the extraction of sensitive session data, including agent-browser cookies to retrieve all browser cookies and agent-browser storage local to access localStorage. If an agent is tricked into visiting a malicious site, these commands could be used to harvest authentication tokens.
[INDIRECT_PROMPT_INJECTION]: The skill is highly susceptible to indirect prompt injection (Category 8).
Ingestion points: The skill reads untrusted data from any website navigated to via agent-browser open and through agent-browser snapshot which parses the DOM/accessibility tree.
Boundary markers: None present; the agent treats the content of the web page as part of its operational context.
Capability inventory: Extensive capabilities including file writing (agent-browser screenshot path.png, agent-browser state save), network interception, and JavaScript execution.
Sanitization: No evidence of sanitization or filtering of the content retrieved from the browser before it is processed by the agent.

agent-browser